ARITHMETICALLY COHEN-MACAULAY BUNDLES ON HYPERSURFACES

ARITHMETICALLY COHEN-MACAULAY BUNDLES ON HYPERSURFACES N. MOHAN KUMAR, A. P. RAO, AND G. V. RAVINDRA Abstract. We prove that any rank two arithmetically Cohen- Macaulay vector bundle on a general hypersurface of degree at least three in P 5 must be split. 1. Introduction Let X P n be a hypersurface of degree d. A vector bundle E on X is called Arithmetically Cohen-Macaulay (ACM for short) if H i (E(k)) = 0 for all k and 0 < i < n 1. By Horrock s criterion [6], this is equivalent to saying that E has a resolution, 0 F 1 F 0 E 0, where the F i s are direct sums of line bundles on P n. If d = 1, E is a direct sum of line bundles (op. cit); the ACM condition is vacuous for n = 1, 2. In this article, we will be interested in ACM bundles of rank two. For n = 3, ACM rank two bundles are ubiquitous [Remark 4]. Hence we will deal with smooth hypersurfaces X of degree d 2 in P n with n 4 and ACM rank two bundles on X which are not direct sums of line bundles of the form O X (k). By the Grothendieck-Lefschetz theorem, these bundles are the same as the indecomposable rank two ACM bundles on X. Our main theorem is Theorem 1.1. Let X P n be a smooth hypersurface of degree d 2. (1) If n 6, any ACM rank two bundle on X is a direct sum of line bundles (Kleppe [8]). (2) (a) If n = 5 and X is smooth, then for any ACM rank two bundle E on X, H 1 (X, E E(k)) = 0 k. In particular, E is rigid. 1991 Mathematics Subject Classification. 14F05. Key words and phrases. Vector bundles, hypersurfaces, Arithmetically Cohen- Macaulay. 1

2 N. MOHAN KUMAR, A. P. RAO, AND G. V. RAVINDRA (b) If n = 5 and X is general (for a dense Zariski open subset of the parameter space of hypersurfaces) of degree d 3, then any ACM bundle of rank two is a direct sum of line bundles. (3) If n = 4 and X is general of degree d 5, then for any ACM rank two bundle E on X, H i (X, E E) = 0 for i = 1, 2. In particular, E is rigid. Remarks: (1) Part 1) of the theorem is known by the work of Kleppe ([8], see Proposition 3.2) where he proves a much more general theorem in this direction. Our proof of this particular case is different and essentially falls out of some of the computations necessary for part 2) of the theorem. (2) If n = 5 and d 2, there certainly exist special smooth hypersurfaces with non-split ACM bundles. Here is a fairly simple way to construct them: Let f, g, h be a regular sequence of homogeneous polynomials. Let a, b, c be non-constant homogeneous polynomials such that these six polynomials have no common zero in P n and such that F = af + bg + ch is homogeneous of degree d 2. Let X be defined by F = 0. Then we have an exact sequence, 0 E O X ( deg f) O X ( deg g) O X ( deg h) I 0, where I is the ideal generated by f, g, h in O X, the map to I is the obvious one and E is the kernel. One easily checks that E is an indecomposable ACM bundle on X of rank two. Any smooth quadric hypersurface in P 5 has a Plücker equation F = 0 and so the above construction applies. If n = 5 for d = 3, 4, 5 and 6, the statement of Theorem 2 b) has been proved already by Chiantini and Madonna [5]. (3) In the case of n = 4, the rigidity statement was proved for quintic threefolds by Chiantini and Madonna [3]. Further, it is known that indecomposable ACM bundles of rank 2 exist for any smooth hypersurface of degree d with 2 d 5 ([1], [2], [7]). One way to see this is to note that any such hypersurface contains a line and hence the construction in remark (2) applies. On the other hand, it was shown by Chiantini and Madonna [4] that such bundles do not exist for a general sextic in P 4 and one expects the same to be true for general hypersurfaces of degree d 6 in which case our result is trivially true. (4) When n = 3, any smooth hypersurface contains a point and hence the construction in remark (2) applies in this case too.

ARITHMETICALLY COHEN-MACAULAY BUNDLES 3 We now give a brief sketch of the proof of (1.1) (2b). Suppose we have a rank two indecomposable ACM bundle E on a hypersurface X of degree d. We show that the module N = H 2 (X, E E(k)) is a non-zero graded cyclic module generated by an element of degree d (Lemma 2.3). In the P 5 case, N is a Gorenstein module and the socle element is in degree 2d 6. Thus we see that N k 0 for all d k 2d 6. By means of a deformation-theoretic argument, we will show that if X is general, then the multiplication map N d N 0 by any g H 0 (X, O X (d)) is zero [Corollary 3.8], which by the cyclicity implies that N 0 = 0. This is a contradiction if d is at least three. 2. Cohomology computations We will work over an algebraically closed field of characteristic zero, though most of the arguments will go through in characteristic not equal to two. We will assume throughout that X is a smooth hypersurface of degree d 2 in P n with n 4. Let E be rank two bundle on X. By the Grothendieck-Lefschetz theorem, Pic P n Pic X is an isomorphism. So by normalising E, we will assume that c 1 (E) = e where e = 0 or 1. We will now assume E is ACM. Then we have a minimal resolution 0 F 1 F 0 E 0, where the F i s are direct sums of line bundles on P n of rank r (E = O X (a) O X (b) for some integers a and b if and only if r = 2). Dualizing the above we get, 0 F 0 F 1 Ext 1 O P n(e, O P n) 0. Applying Hom OP n(e, ) to the exact sequence, we see that, This implies that F1 resolution as 0 O P n O P n(d) O X (d) 0, Ext 1 O P n(e, O P n) = E (d) = E(d e). = F 0 (d e). Thus we can rewrite our minimal (1) 0 F 1 = F 0 (e d) φ F 0 E 0 In fact by [2], the map φ can be chosen to be skew-symmetric though we will not use this fact. By restricting to X, we get an exact sequence, (2) 0 E( d) F 1 O X = F 1 φ F0 O X = F 0 E 0

4 N. MOHAN KUMAR, A. P. RAO, AND G. V. RAVINDRA Let the image of φ be denoted by G. Then G is a vector bundle of rank r 2 on X. Taking exterior powers of φ in (1), we have an exact sequence (3) 0 Λ 2 F 1 Λ 2 F 0 F 0 for some cokernel sheaf F. Lemma 2.1. We have an exact sequence 0 L F F O X = F 0 where L is a line bundle on X and F is a vector bundle of rank 2r 3 on X which fits into a natural exact sequence 0 E G F O X (e) = Λ 2 E 0. Proof. We certainly have an exact sequence 0 IF F F 0 where I is the ideal sheaf defining X in P n. It suffices to check that IF is a line bundle on X and F is a vector bundle of rank 2r 3 for the first part of the lemma. Both statements are local. Clearly F is set-theoretically supported only along X. By localising we may assume that (1) looks like, 0 F 1 F 0 E 0 where the F i s are free of rank r and the matrix of the map F 1 F 0 is the diagonal matrix (x, x, 1, 1,..., 1) where x = 0 defines X. Then the matrix in (3) is given by the diagonal matrix (x 2, x,... x, 1, 1,..., 1) where we have one x 2, 2(r 2) x s and the rest 1 s. The claim about IF and F follows easily from this. To see the final exact sequence, we restrict (3) to X to get From the exact sequence, we note that Λ 2 F 1 Λ 2 F 0 F 0. 0 G F 0 E 0 im (Λ 2 F 1 Λ 2 F 0 ) = im (Λ 2 G Λ 2 F 0 ). This leads to the natural diagram,

ARITHMETICALLY COHEN-MACAULAY BUNDLES 5 0 0 Λ 2 G = Λ 2 G 0 G Λ 2 F 0 Λ 2 E = O X (e) 0 0 E G F O X (e) 0 0 0 where G is defined by the diagram. Next we note the vanishing of various cohomologies. From (1) and (3) we have (4) H i (E( )) = H i (F( )) = H i (L( )) = 0, 1 i n 2 This implies from the lemma above that (5) H i (F( )) = 0, 1 i n 3 Tensoring the exact sequence, 0 E( d) F 1 G 0 with E and taking cohomologies, we get, (6) H i (E G(k)) = H i+1 (E E( d + k)) k, 1 i n 3 Similarly, tensoring the exact sequence 0 G F 0 E 0 with E and taking cohomologies, we get, (7) H i (E E(k)) = H i+1 (E G(k)) k, 1 i n 3 From the exact sequence in the lemma, using (4),(5) we get, (8) H i (E G(k)) = 0, 2 i n 3 Lemma 2.2. The vector bundle E is a direct sum of line bundles if and only if H 2 (E E( d)) = 0. Proof. If E is a direct sum of line bundles, this vanishing is clear. So assume the vanishing. From (6) we see that Ext 1 (E, G) = H 1 (E G) = 0 and thus we see that the exact sequence 0 G F 0 E 0 splits. Since F 0 is a direct sum of line bundles, we see that so is E. Corollary 2.3. If E is an indecomposable bundle, then the finite length module k H 1 (E G(k)) is a non-zero cyclic module generated by an element of degree e.

6 N. MOHAN KUMAR, A. P. RAO, AND G. V. RAVINDRA Proof. By lemma 2.1, we have an exact sequence, 0 E G F O X (e) 0. Taking cohomologies and using (5), we get the fact that the module is cyclic generated by an element of degree e. If it is zero then H 1 (E G( e)) = H 1 (E G) = 0 and so by (6) H 2 (E E( d)) = 0. By the previous lemma, the bundle would then have to be split. 3. Deformation criteria for acm vector bundles We have already noted that for n 5, given any degree d 2, there exists a smooth hypersurface of degree d and a non-split ACM bundle of rank 2 on the hypersurface. So, in this section, we analyze the situation, when there is such a vector bundle on a general hypersurface X of degree d. We will start with some results on vector bundles on families of varieties. We will not prove the most general results in this direction, but just what we need. Most of the arguments are similar to those used in the construction of quot schemes and are well-known to experts. We are really interested in Corollary 3.5 and much of what follows consists of technical results to achieve it. Let us start by fixing some notation. All schemes will be of finite type over the base field. Let p : X S be a flat projective morphism. If q : T S is any morphism, we will denote by X T = X S T, the fiber product and p : X T T, q : X T X, the natural S-morphisms. We start with an elementary lemma, whose proof we omit. Lemma 3.1. Let X, S, T be as above. Let V X be any subset such that q (X T ) V =. Then, q (X T ) p 1 p(v ) =. Proposition 3.2. Let p : X S be a flat projective morphism. Let F 1, F 2 be two vector bundles on X with H 1 (X s, F 2 F 1 Xs ) = 0 s S. Let m 0 be an integer.then there exists a scheme q : S S and an exact sequence, q F 2 q F 1 G 0 where G is a rank m vector bundle on X S. Furthermore, one has the following universal property: for any reduced S-scheme T with structure morphism f : T S and an exact sequence f F 2 f F 1 G 0

ARITHMETICALLY COHEN-MACAULAY BUNDLES 7 with G a rank m vector bundle on X T, there exists an S-morphism T S such that the second sequence is just the pull back of the first by the induced morphism X T X S. Proof. The hypothesis on H 1 ensures that E = p (F2 F 1 ) is a vector bundle on S. Let H = A(E ) q S. Then q E has a section and H is universal with this property. We have the fiber product diagram, By the flatness of q, we have, X p p H q S p q (F 2 X H q F 1 ) = q E and so q (F2 F 1 ) has a section. Thus on X H we get the universal map q F 2 q F 1. Let G be the cokernel. By semi-continuity, the points x X H such that dim k(x) G k(x) < m constitute an open set, which we denote by V. Since p is flat p (V ) is open. Let H H be the closed subset with the reduced scheme structure which is the complement of p (V ). Thus, replacing H with H, we get an exact sequence, where G has the property that at every point x X H, dim k(x) G k(x) m. Again, the set of points x X H such that dim k(x) G k(x) > m is a closed subset, say Z. Since p is proper, we may take S H to be the open set which is the complement of the closed subset p (Z). It is clear that on X S we have an exact sequence as claimed. We need to check the universal property. So, let f : T S be as in the proposition with the exact sequence as mentioned. Let us denote by f : X T X and p : X T T the corresponding maps. The existence of the map f F 2 f F 1 gives a section of p f (F 2 F 1 ) which is equal to f E by semi-continuity. This implies that we have a morphism g : T H over S and the exact sequence q E 2 q E 1 G 0 on X H pulls back to the one on X T via the induced map g : X T X H. Since G is a rank m vector bundle, we see that g (X T ) V = and thus we see that by lemma 3.1, g (X T ) p 1 p (V ) =. Since T is reduced, this implies that g factors through H. Next, we notice that g (X T ) Z = and by another application of lemma 3.1, we are done.

8 N. MOHAN KUMAR, A. P. RAO, AND G. V. RAVINDRA Given a rank two (non-split) ACM bundle E on X, we rewrite (1) with F 0 = r i=1 O P n(a i ), a 1 a 2 a r and r > 2 to get 0 r i=1o P n( a i + e d) φ r i=1o P n(a i ) E 0 As before, we assume that e = 0 or e = 1. Lemma 3.3. When n 4, for fixed d there are only finitely many possibilities for a := (a 1,, a r ). Proof. Since sequence (1) is minimal, we see that all entries of φ must be at least of degree 1 and thus deg det φ r. But this determinant is just the square of the equation defining our hypersurface and thus it must be 2d. So r 2d and is bounded. From the fact that φ is an inclusion, we see that a r + e d a 1. Thus, if we show that a 1 is bounded above, then it will follow that a r is bounded below and then we will have only finitely many possibilities for the a i s. For this we proceed as follows. Let a = a 1. Since (1) is minimal, we get an inclusion O X (a 1 ) E such that the quotient is torsion free. Thus, we get an exact sequence, 0 O X (a) E I(e a) 0 where I is the ideal sheaf of a codimension two subscheme Z X. This implies ω Z = OZ (e 2a + d n 1) where ω Z is the the canonical bundle of Z. Let π : Z P n 3 be a general projection, so that π is finite. Then π ω Z = Hom(π O Z, O P n 3( n + 2)). Since O P n 3 is a direct summand of π O Z, H 0 (π ω Z (n 2)) 0 and thus H 0 (O Z (e 2a + d 3)) 0. Since n 4, and E is ACM we see that H 1 (I( )) = 0 and thus H 0 (O X (k)) H 0 (O Z (k)) is onto. Since X is a smooth hypersurface, H 0 (O X (k)) = 0 for k < 0. Thus we see that e 2a + d 3 0 or a (e + d 3)/2. Theorem 3.4. Let P denote the parameter space of all degree d smooth hypersurfaces in P n with n 4 and X P n P the universal hypersurface. Assume that there exist rank two ACM non-split bundles on hypersurfaces corresponding to a Zariski dense subset of P. Then there exists a scheme P, a dominant morphism P P and a rank two bundle over X P P which is ACM and non-split for any point in P. Proof. By lemma 3.3, for any rank two non-split ACM bundle on a hypersurface X of degree d, we have finitely many choices for r, e and the a i s. Let us fix one of these. Then for all vector bundles E with these invariants, we have a presentation, F 0 (e d) O X F 0 O X E 0.

ARITHMETICALLY COHEN-MACAULAY BUNDLES 9 Thus, we consider the flat projective morphism p : X P and the vector bundles F0 (e d), F 0 pulled back to X, which we denote by the same name. Notice that the vanishing condition on first cohomology in Proposition 3.2 holds if n 3. Thus we get a scheme q : S = S (e, r, a i ) P as in the proposition for m = 2. Let G be the corresponding rank two bundle on X S. We have a closed subset S S where the map F0 (e d) F 0 is minimal. We may restrict to S and let p : X S S be the structure map. By the relative version of Serre vanishing [9],there exists an integer m 0 such that for all m m 0 and all i > 0, R i p G(m) = 0. Since G is a vector bundle and p is flat, by repeated application of semi-continuity theorems [see for example, page 41, [9]], one sees that H i (X s, G(m) Xs ) = 0 for all i > 0 and all m m 0. By duality this is also true for all i < n 1 and all m m 1 for a suitable m 1. Thus we see that there are only finitely many integers k such that H i (X s, G(k) Xs ) 0 for some s S and some i with 0 < i < n 1. Since the set of s S such that H i (X s, G(k) Xs ) 0 for fixed i, k is a closed subset, we see that there is a well-defined closed subset Z S such that H i (X s, G(k) Xs ) 0 for some k and some i with 0 < i < n 1 if and only if s Z. If we let T = S Z, we see that on X T, the bundle G has the property that on each fibre over T, it is ACM and non-split. Let P = T, the union taken over all possible choices of r > 2, e and the a i s. Thus, we get rank two non-split ACM bundles on all fibres of X P P. To prove that P P is dominant, it suffices to show that the image of this map contains a dense set. Let x P be a point such that X x supports a rank two non-split ACM bundle, say E. Let r, e, a i be the corresponding invariants. Then by the universal property of Proposition 3.2, we see that there exists a point y S = S (r, e, a i ) such that q(y) = x and the pull back of the corresponding G is this bundle E. By minimality of our resolution, we see that y S. Since E is ACM, we see that y T. This completes the proof. For a hypersurface X P n of degree d, the total infinitesimal deformation X A of X in P n is contained in P n Spec A where A = k V, V = H 0 (O X (d)) and V 2 = 0 in the ring A. The following corollary uses characteristic zero. Corollary 3.5. Assume that we have non-split rank two ACM bundles on a general hypersurface of degree d in P n with n 4. Then for a general hypersurface X of degree d there exists a rank two bundle E on X A such that E X is a non-split ACM bundle.

10 N. MOHAN KUMAR, A. P. RAO, AND G. V. RAVINDRA Proof. By theorem 3.4, under the hypothesis, we have a finite type scheme P mapping dominantly to P and a rank two bundle on the universal hypersurface in P n P which is non-split and ACM on each fiber. Since P is integral, by replacing P, we may clearly assume that it is integral. Since the map is dominant, we may take a generic multisection and thus assume that P P is generically finite. Replacing again P by a suitable open set, we may assume that the map is etale since the characteristic is zero. If x P corresponding to X P n is the image of y P, then the tangent spaces at x and y are isomorphic. Now noting that V is the tangent space at x to P, we are done. For the next theorem, we will need the following elementary lemma from homological algebra, whose proof (which we omit) follows from the construction of push-outs. Lemma 3.6. Let 0 A i B j j 0 A i B be a push-out diagram where A, A, B, B are sheaves. Then i splits if and only if there exists a homomorphism α : B A such that α i = j Before we state the next theorem, let us fix some notation. Let X be a smooth hypersurface of degree d in P n with n 3 defined by F = 0 and E an ACM bundle on X of arbitrary rank. Let G H 0 (O P n(d)) be such that the image g V = H 0 (O X (d)) is non-zero. Let B = k[ɛ] with ɛ 2 = 0. The map g : V kɛ defines a natural surjective ring homomorphism A B where A is as above. Let X ɛ P n Spec B be the corresponding hypersurface defined by F ɛg = 0. In the theorem we will assume that there is a bundle E on X ɛ such that the restriction of E to X X ɛ is E. As always, we have a resolution, 0 F 1 φ F0 E 0, with the F i s direct sums of line bundles on P n. This gives as before, by restriction to X, a long exact sequence, and by splitting it to short exact sequences, an exact sequence, 0 E( d) F 1 G 0 for some bundle G on X as in (2). Let us denote by ζ the corresponding extension class in Ext 1 O X (G, E( d)) = H 1 (G E( d)).

ARITHMETICALLY COHEN-MACAULAY BUNDLES 11 Theorem 3.7. With the notation and assumptions as above, under the natural map H 1 (G E( d)) g H 1 (G E)) the image of ζ is zero. Proof. Since E is flat over k[ɛ], we have, We get an exact sequence E = ɛe = E/ɛE. 0 E = ɛe i E π E = E/ɛE 0. Let p : X ɛ P n be the projection, which is clearly a finite map. Taking the direct image under p of the above exact sequence and noting that p restricted to X X ɛ is a closed embedding in P n, we get (9) 0 E i p E = F π E 0. We want to interpret multiplication by F on F. We know that F ɛg annihilates E, and so multiplication by F is the same as multiplication by ɛg. But multiplication by ɛ is just the composite i π. Thus we see that multiplication by F is the composite (10) F π E g E(d) i F(d), using the fact that i g = g i : E F(d). The exact sequence (9) gives an element η in Ext 1 O P n(e, E). From the exact sequence we get 0 F 1 F 0 E 0, H 0 (F 1 E) Ext 1 O P n(e, E) H 1 (F 0 E). Since E is ACM and n 3 the last term is zero. Thus we can lift η to an α H 0 (F1 E) and we get the diagram 0 F 1 φ F0 E 0 α β 0 E i F π E 0 Multiplication by F gives the following commutative diagram F 0 ( d) β F( d) F F 0 β F F

12 N. MOHAN KUMAR, A. P. RAO, AND G. V. RAVINDRA Since F annihilates E, the top row factors as F 0 ( d) ψ F 1 φ F0 for a suitable map ψ. Using (10) we get the following diagram, which we will show is in fact commutative. F 0 ( d) ψ F 1 φ F0 β α β π F( d) E( d) E F We have igπβ = F β = βf = βφψ = iαψ. Since i is an inclusion, this implies, gπβ = αψ, proving commutativity. Restricting this diagram to X, we get g i F 0 ( d) ψ F 1 φ F0 β α β F OP n O X ( d) π E( d) g E i F OP n O X The image of ψ is just E( d) and thus we get a commutative diagram E( d) F 1 α E( d) g E The extension class ζ corresponds to the top row of the following commutative diagram 0 E( d) F 1 G 0 g α E and by lemma 3.6 under the natural map Ext 1 (G, E( d)) g Ext 1 (G, E), ζ goes to zero. Therefore under the map ζ goes to zero. H 1 (G E( d)) g H 1 (G E) Corollary 3.8. Assume the rank of E is two, n 4 and that as before, E can be deformed in the direction of g. Then the natural map is zero. H 2 (E E( d)) g H 2 (E E)

ARITHMETICALLY COHEN-MACAULAY BUNDLES 13 Proof. From theorem 3.7 under the map, H 1 (G E( d)) g H 1 (G E), ζ goes to zero. By (2), we have G = G(d e). Thus under the map H 1 (G E( e)) g H 1 (G E(d e)), ζ goes to zero. If E is indecomposable, H 1 (G E( e)) is one dimensional with ζ as basis element by corollary 2.3. From (6) we have, H 1 (G E( e + k)) = H 2 (E E( d + k)). Thus we get that the map is zero. H 2 (E E( d)) g H 2 (E E) 4. Proof of theorem 1.1 Proof of Theorem 1.1 (1). For n 6, H 2 (E E( d)) = H 3 (E G( d)) by (7). This group is zero by (8). The proof now follows from lemma 2.2. Proof of Theorem 1.1 (2) a). For n = 5, using (7) with i = 1, we have This group is zero by (8). H 1 (X, E E(k)) = H 2 (X, E G(k)). Proof of Theorem 1.1 (2) b). The proof is by contradiction. Assume that a general hypersurface X of degree d 3 has an indecomposable ACM bundle of rank two. From corollaries 3.5 and 3.8, we see that, H 2 (E E( d)) g H 2 (E E) is zero for any g H 0 (O X (d)). By corollary 2.3 and (6), we know that the graded module N = k H 2 (E E(k)) is cyclic and generated by a non-zero element ζ in degree d. Thus N 0 consists of multiples of ζ by elements g H 0 (O X (d)). Since these are zero we get N i = 0 for i 0. As n = 5, we have by Serre duality H 2 (E E( d)) = H 2 (E E(2d 6)). Hence N 2d 6 0. If d 3, then 2d 6 0 and this contradiction proves the result.

14 N. MOHAN KUMAR, A. P. RAO, AND G. V. RAVINDRA Proof of Theorem 1.1 (3). Assume that a general hypersurface X of degree d 5 has an (indecomposable) ACM bundle of rank two. By the same arguments as in the proof above, the graded module N has N i = 0 for i 0 and thus H 2 (E E) = 0. By Serre duality, we have H 1 (E E) = H 2 (E E(d 5)) = N d 5 and hence is zero for d 5. References [1] Adler, Allan; Ramanan, S. Moduli of abelian varieties, Lecture Notes in Mathematics, 1644. Springer-Verlag, Berlin, 1996. [2] Beauville, Arnaud. Determinantal hypersurfaces, Dedicated to William Fulton on the occasion of his 60th birthday. Michigan Math. J. 48 (2000), 39 64. [3] Chiantini, Luca ; Madonna, Carlo. ACM bundles on a general quintic threefold, Dedicated to Silvio Greco on the occasion of his 60th birthday (Catania, 2001). Matematiche (Catania) 55 (2000), no. 2, 239 258 (2002). [4] Chiantini, Luca ; Madonna, Carlo. A splitting criterion for rank 2 bundles on a general sextic threefold, Internat. J. Math. 15 (2004), no. 4, 341 359. [5] Chiantini, L.; Madonna, C. K. ACM bundles on general hypersurfaces in P 5 of low degree. Collect. Math. 56 (2005), no. 1, 85 96. [6] Horrocks, G. Vector bundles on the punctured spectrum of a local ring, Proc. London Math. Soc. (3) 14 (1964) 689 713. [7] Iliev, A.; Markushevich, D. Quartic 3-fold: Pfaffians, vector bundles, and halfcanonical curves, Michigan Math. J. 47 (2000), no. 2, 385 394. [8] Kleppe, Hans. Deformation of schemes defined by vanishing of Pfaffians, J. Algebra 53 (1978), no. 1, 84 92. [9] Mumford, David. Lectures on curves on an algebraic surface, Annals of Mathematics Studies, No. 59 Princeton University Press. Department of Mathematics, Washington University in St. St. Louis, Missouri, 63130 E-mail address: kumar@wustl.edu URL: http://www.math.wustl.edu/~ kumar Louis, Department of Mathematics, University of Missouri-St. Louis, St. Louis, Missouri 63121 E-mail address: rao@arch.umsl.edu Department of Mathematics, Washington University in St. St. Louis, Missouri, 63130 E-mail address: ravindra@math.wustl.edu Louis,